gal-tab/sir_albert/decide

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Decision memory: record a call once, auto-apply it every time the same question recurs — never re-decide what's already settled. Use when the user says "/decide", "record this decision", "we decided", "did we decide X before", "log this call", or "is this already decided".

Install

npx vskill@latest install gal-tab/sir_albert/decide

SKILL.md

Decide

Boot from os/PREAMBLE.md.

Purpose: stop re-deciding recurring calls. Every settled question gets one authoritative record in os/state/decisions.jsonl. When the same question surfaces again, the prior decision is stated and applied automatically — no re-ask.


0. Check before asking — always

Before prompting the user about anything that looks like a policy, threshold, or recurring call, scan os/state/decisions.jsonl for a matching question_id:

bash
grep '"question_id"' os/state/decisions.jsonl

Or for a specific slug:

bash
grep 'suppress-existing-customers' os/state/decisions.jsonl

If a match exists and the scope / project still applies, do not ask — auto-apply (see §2).


1. Store — recording a new decision

Append exactly one JSON object (no newline inside the object, blank line between entries is fine) to os/state/decisions.jsonl:

json
{"ts":"2026-07-31","question_id":"suppress-existing-customers","choice":"yes","rationale":"Re-engaging existing customers from acquisition campaigns inflates CAC and confuses attribution.","project":"*","scope":"gtm"}

Field rules:

FieldFormatNotes
tsYYYY-MM-DDDate the decision was made
question_idkebab-case-slugStable identifier — never change once written
choiceshort stringThe answer: "yes", "no", "last-touch", "≥1000", etc.
rationale1–2 sentencesThe why — future-you needs this, not just the what
projectstring or "*"Specific project slug or "*" for org-wide
scopestringDomain label: "gtm", "data", "ab-testing", "consent", etc.

Append command (never overwrite, never delete):

bash
echo '{"ts":"YYYY-MM-DD","question_id":"<slug>","choice":"<choice>","rationale":"<why>","project":"<project-or-*>","scope":"<scope>"}' >> os/state/decisions.jsonl

Confirm the append by reading back the last line:

bash
tail -1 os/state/decisions.jsonl

2. Auto-apply — when a prior decision matches

When a question arises that maps to an existing question_id (and project is "*" or matches the current project, and scope is relevant):

  1. Do not ask the user to re-decide.
  2. State the prior decision inline, in this format:

> Prior decision (2026-06-01): suppress existing customers = yes, because re-engaging them inflates CAC and confuses attribution — auto-applied.

  1. Proceed with the choice already applied.

If the user wants to override, they say so explicitly — then record a new entry with the updated choice and ts. The old entry stays (append-only log).


3. Recurring examples

These question_ids cover the most common recurring calls. Check for them before asking.

question_idTypical choiceScope
suppress-existing-customersyesgtm
ab-test-min-audience-size≥1000 per variantab-testing
attribution-model-paid-sociallast-touchdata
consent-container-ownermonday-marketingconsent

4. Lookup mode (/decide without a new decision)

If the user asks "did we decide X?" or "is this already decided?", run the grep from §0, then:

  • Found: surface the full JSON object, formatted as a human-readable summary (ts, choice, rationale), and note whether it auto-applies to the current context.
  • Not found: say "No recorded decision for this question. Want me to record one now?" — then follow §1 if yes.

Completion report

After any store or lookup, confirm:

Decision logged: <question_id> → <choice> (os/state/decisions.jsonl, appended)

or

Decision found: <question_id> → <choice> (recorded <ts>) — auto-applied.

Do NOT commit decisions.jsonl unless the user explicitly asks.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/gal-tab/sir_albert/decide/badge)](https://verified-skill.com/skills/gal-tab/sir_albert/decide)

Something wrong with this skill or its page? Tell us.